Snails on Oceanic Islands: Testing the ARTICLE General Dynamic Model of Oceanic Island Biogeography Using Linear Mixed Effect Models Robert A

Home , Amastridae, Enidae, Hemicycla, Lauriidae, Lyropupa, Partulidae

Journal of Biogeography (J. Biogeogr.) (2013) 40, 117–130

ORIGINAL Snails on oceanic islands: testing the ARTICLE general dynamic model of oceanic island biogeography using linear mixed effect models Robert A. D. Cameron1,2, Kostas A. Triantis3,4,5*, Christine E. Parent6, François Guilhaumon3,7, Marıá R. Alonso8, Miguel Ibań˜ez 8, Antońio M. de Frias Martins9, Richard J. Ladle4,10 and Robert J. Whittaker4,11

1Department of Animal and Plant Sciences, ABSTRACT University of Sheffield, Sheffield, S10 4TN, Aim We collate and analyse data for land snail diversity and endemism, as a UK, 2Department of Zoology, The Natural means of testing the explanatory power of the general dynamic model of oce- History Museum, London, SW7 5BD, UK, 3Azorean Biodiversity Group, University of anic island biogeography (GDM): a theoretical model linking trends in species Azores, 9700-851, Angra do Heroı´smo, immigration, speciation and extinction to a generalized island ontogeny. 4 Terceira, Azores, Portugal, Conservation Location Eight oceanic archipelagos: Azores, Canaries, Hawaii, Gala´pagos, Biogeography and Macroecology Programme, Madeira, Samoa, Society, Tristan da Cunha. School of Geography and the Environment, University of Oxford, Oxford, OX1 3QY, UK, Methods Using data obtained from literature sources we examined the power 2 5Department of Ecology and Taxonomy, of the GDM through its derivative ATT model (i.e. diversity met- 2 Faculty of Biology, National and Kapodistrian ric = b1 + b2Area + b3Time + b4Time ), in comparison with all the possible University, Athens, GR-15784, Greece, simpler models, e.g. including only area or time. The diversity metrics consid- 6Section of Integrative Biology, University of ered were the number of (1) native species, (2) archipelagic endemic species, Texas, Austin, TX 78712, USA, 7‘Rui and (3) single-island endemic species. Models were evaluated using both log- Nabeiro’ Biodiversity Chair, CIBIO – transformed and untransformed diversity data by means of linear mixed effect Universidade de E´vora, Casa Cordovil, 7000- models. For Hawaii and the Canaries, responses of different major taxonomic 8 890, E´vora, Portugal, Departamento de groups were also analysed separately. Biologıá Animal, Universidad de La Laguna, 2 E-38206 La Laguna, Tenerife, Canaries, Results The ATT model was always included within the group of best models Spain, 9CIBIO-Açores, Department of Biology, and, in many cases, was the single-best model and was particularly successful University of the Azores, 9501-801 Ponta in fitting the log-transformed diversity metrics. In four archipelagos, a hump- Delgada, Azores, Portugal, 10Institute of shaped relationship with time (island age) is apparent, while the other four Biological Sciences and Health, Federal archipelagos show a general increase of species richness with island age. In University of Alagoas, AL, Brazil, 11Centre for Hawaii and the Canaries outcomes vary between different taxonomic groups. Macroecology, Evolution and Climate, Main conclusions The GDM is an intentionally simplified representation of Department of Biology, University of Copenhagen, DK-2100, Copenhagen, environmental and diversity dynamics on oceanic islands, which predicts a Denmark simple positive relationship between diversity and island area combined with a humped response to time. We find broad support for the applicability of this model, especially when a full range of island developmental stages is present. However, our results also show that the varied mechanisms of island origins and the differing responses of major taxa should be taken into consideration when interpreting diversity metrics in terms of the GDM. This heterogeneity is reflected in the fact that no single model outperforms all the other models for all datasets analysed. *Correspondence: Kostas A. Triantis, Azorean Keywords Biodiversity Group, University of Azores, 9700- General dynamic model, island biogeography theory, island evolution, land 851 Angra do Heroı´smo, Terceira, Azores, Portugal. snails, mixed effect models, model selection, oceanic islands, speciation, spe- E-mail: [email protected] cies diversity dynamics.

ª 2012 Blackwell Publishing Ltd http://wileyonlinelibrary.com/journal/jbi 117 doi:10.1111/j.1365-2699.2012.02781.x R. A. D. Cameron et al.

Martins, 2005; Alonso et al., 2006; Parent & Crespi, 2006, INTRODUCTION 2009). Many land snail species are endemics with tiny geo- Islands are prime reservoirs of land snail diversity graphical ranges, a feature that is even more apparent in snails inhabiting oceanic islands (Solem, 1984). In this paper Alan Solem (1984, p. 9) we examine the patterns of snail species diversity and ende- The biotas of oceanic islands have provided ideal material for mism of oceanic islands, within the context of the GDM, studying ecological and evolutionary processes and their analy- and we test the predictions and the generality of the mathe- sis has profoundly influenced the development of subjects matical models arising from the GDM, using first, the overall ranging from species formation to biogeography (Whittaker & faunas, and second, certain taxonomic subsets. We do so Fernańdez-Palacios, 2007). Whittaker et al. (2008) recently using linear mixed models, an analytical approach pioneered developed a model, the general dynamic model of oceanic in this context by Bunnefeld & Phillimore (2012). island biogeography (hereafter GDM), to account for the development of species richness on oceanic islands. Building MATERIALS AND METHODS on the dynamic equilibrium theory of MacArthur & Wilson (1967), the GDM explicitly places the outcome of the processes The archipelagos of immigration, speciation and extinction in a temporal context dictated by the geological development of the islands. Data from 56 islands from eight oceanic archipelagos were The GDM is based on three premises: (1) that emergent selected based on the availability of reliable faunal lists, and properties of island biotas are a function of predictable an estimated age of origin (maximum age) of each of the trends in rates of immigration, speciation and extinction; (2) islands (see Appendix S1 in Supporting Information). Oce- that evolutionary dynamics predominate in large, remote anic islands are generally considered to be those that have islands/archipelagos; and (3) that oceanic islands are rela- formed over oceanic crust and that have never been con- tively short-lived landmasses showing a characteristic nected to continental landmasses (Whittaker & Fernańdez- humped trend in carrying capacity over their life span. The Palacios, 2007): they are typically relatively short-lived land- model predicts that several key diversity metrics (e.g. the masses and relatively few last longer than a few million years number of native species, the number of archipelagic ende- before subsiding and/or eroding back into the sea. Despite mic species etc.) should show a general hump-shaped trend the relative simplicity of the geological history of some oce- over time, driven largely by changing carrying capacity over anic island groups, e.g. Hawaii, the dynamics of most archi- the life cycle of the island itself. As we are unable to follow pelagos are more complex than assumed within the GDM the behaviour of a single island over millions of years, Whit- (e.g. Courtillot et al., 2003; Neall & Trewick, 2008). Even for taker et al. (2008) suggest that certain predictions could be hotspot archipelagos with a more or less linear arrangement tested by examining contemporary data from across the dif- of islands, at least three distinct types have been identified, ferent aged islands of an oceanic archipelago. This, in turn, based largely on the origin of the plumes (Fig. 4 in Courtil- generates an expectation of a positive relationship between lot et al., 2003). Nevertheless, for the majority of the islands these diversity metrics and island area combined with a para- considered here, the age of origin (maximum age) is more bolic relationship with time, i.e. diversity = b1 + b2Area + or less agreed upon (below), and thus we use maximum age 2 b3Time + b4Time (where ‘Time’ is time elapsed since island for all the analyses here (Table 1). emergence and ‘Area’ is island area), provided that the archi- Island age data were derived as follows: (1) Hawaiian pelago(s) concerned display a full range of island develop- Islands: Clague (1996). (2) Gala´pagos Islands: Geist (1996 mental stages. As this is not always the case, the relationship and unpublished data). (3) Azores: although Johnson et al. with time may vary according to the extent of the geological (1998) suggest an age of just 0.8 Ma for Saõ Miguel, we use ages involved, from positive, to hump-shaped, or negative. 4.01 Ma and the other ages adopted by Borges et al. (2009) Thus, fits employing this ATT2 model framework can take for the rest of the archipelago in formal analyses. (4) Madei- different forms, such as simple linear area–time dependence ran group: Geldmacher et al. (2005). (5) Canary Islands: (i.e. without the quadratic term of time), or a negative Carracedo et al. (2002). In the case of Gran Canaria, an age dependence on time (i.e. with a positive relationship with of c. 3.5 Ma has been used in some previous analyses (see area and a negative relationship with time) (see Fig. 1; Whit- Whittaker et al., 2008 and discussion therein) based on the taker et al., 2008, 2010; Borges & Hortal, 2009; Cardoso hypothesis of near-complete sterilization in the catastrophic et al., 2010; Triantis et al., 2010). Relationships may also be Roque Nublo ash flow (Marrero & Francisco-Ortega, 2001). influenced by the specific ecological requirements and dis- However, Anderson et al. (2009) demonstrate that this persal powers of the taxa studied (e.g. cave-adapted Azorean hypothesis is implausible, hence we use the maximum sub- arthropods compared to other arthropods; see Borges & aerial age of Gran Canaria, i.e. c. 14.5 Ma (Carracedo et al., Hortal, 2009). 2002; for discussion see Fernańdez-Palacios et al., 2011). (6) Land snails are among the better known and studied Samoan Islands: Workman et al. (2004), and Neall & Tre- groups of invertebrates within oceanic archipelagos (Solem, wick (2008). (7) Society Islands: Clouard & Bonneville 1984, 1990; Cameron et al., 1996; Cowie, 1996; Chiba, 1999; (2005). (8) Tristan da Cunha: Ryan (2009).

118 Journal of Biogeography 40, 117–130 ª 2012 Blackwell Publishing Ltd Land snails on oceanic islands

Figure 1 The three different forms of the species–area–time relationship for oceanic island groups predicted within the context of the general dynamic model of oceanic island biogeography (Whittaker et al., 2008, 2010; Triantis et al., 2010). The ﬁrst and the third are described by simple log (Area)–Time relationship (AT; with a positive and negative relationship with time for the ﬁrst and the third model, respectively), and a log(Area)–Time–Time2 model (ATT2) for the second (adapted from Triantis et al., 2010).

da Cunha: Holdgate (1965) and Preece & Gittenberger Sources and treatment of snail data (2003). Our sources of data on snail faunas were as follows: Many oceanic island snail faunas have suffered extensive (1) Hawaiian Islands: Cowie et al. (1995) and Cowie (1995), extinctions due to human activity (Solem, 1990; Cowie, updated by reference to Pokryszko (1997) for Lyropupa. (2) 1995, 2001; Coote & Loe`ve, 2003). We have included Gala´pagos Islands: Dall & Ochsner (1928), Smith (1966), described species extinguished by such activity, but we can- Coppois (1985), Parent & Crespi (2006) and references not know about species that became extinct before they had therein. (3) Azores: Cunha et al. (2010), updated with been described, or about segregations that might have unpublished data of A.M.F. Martins, R.A.D. Cameron and B. resulted had modern techniques been available at the time. M. Pokryszko. (4) Madeira group: Seddon (2008) with some Nevertheless, detailed studies of island mollusc faunas started corrections and minor modiﬁcations (following Goodfriend earlier than for most invertebrate groups, and early invento- et al., 1996; Cameron et al., 2007). (5) Canary Islands: ries, including species now extinct, are remarkably complete Nu´ n˜ez & Nu´ n˜ez (2010), updated by Vega-Luz & Vega-Luz (Seddon, 2008). Furthermore, in each case we have excluded (2008), Holyoak & Holyoak (2009), Neiber et al. (2011), and from the analysis species thought to have been introduced. unpublished data of M.R. Alonso and M. Iba´n˜ez. (6) Fossil evidence for early occurrence of non-endemic snails is Samoan Islands: Cowie (1998) with additional records from available in some cases; in others the judgement of the local Cowie (2001) and Cowie et al. (2002). (7) Society Islands: workers has been used. All slugs have been excluded because Peake (1981) provided overall species numbers, but did not nearly all are introduced (e.g. for Hawaii, see Cowie et al., discriminate introduced species. Thus, we use total number 1995). of species for this group of islands in our analyses. We have We have generally followed the taxonomic status as given updated this list adding the recently described species from in the source publication, considering only full species. For Gargominy (2008). Species lists are available only for the each dataset, we compiled and recorded three diversity met- native family Partulidae (Coote & Loe`ve, 2003). (8) Tristan rics (D), i.e. number of native species/species richness (SR),

Table 1 Properties of the oceanic island systems included in the analyses on land snail diversity and endemism. For a full list of data sources, see text.

Island group No. of islands/units* Area (km2) Total area (km2) Elevation (m) Geological age (Ma)

Azores 9 (9) 17–750 2324 2351 0.25–8.12 Canary 7 (6) 278–2058 7601 3711 0.70–20 Gala´pagos 10 (3) 4.99–4588 7847 1707 0.30–6.3 Hawaii 10 (4) 0.2–10,433 16397 4205 0.60–23.4 (All) 0.60–4.7 (Large) Madeira 3 15–740 795 1860 4.60–14 Samoa 7 (4) 0.3–1717.6 3049 1858 0.40–3.2 Society 6 (6) 38–1000 1486 2231 1.19–3.6 Tristan da Cunha 4 (4) 4–96 179 2060 0.20–18

*Number of islands and island units as used for the two analyses of the data followed (see text). For the archipelago of Hawaii ‘(All)’ refers to the maximum geological age for all the 10 islands considered herein, and ‘(Large)’ for the maximum age of the largest islands only (see Appendix S3).

Journal of Biogeography 40, 117–130 119 ª 2012 Blackwell Publishing Ltd R. A. D. Cameron et al. number of archipelagic endemic species (nEnd), and the and comparing regression parameters for the different levels of number of single-island endemic species (nSIE). The nSIE is the factor (e.g. different archipelagos). Considering a factor as a simple metric indicative of evolutionary dynamics that a fixed effect leads to the estimation of different regression reflects the outcome of in situ speciation, extinction and parameters for continuous variables, such as area, time or ele- migration within the islands of an archipelago. Additionally, vation for each level of the factor. When studying the fixed for Hawaii and the Canaries, we subdivided faunas into effect of a factor, the implicit assumption is made that the lev- major taxonomic groups. els of the factor considered in the analysis are exhaustive (e.g. We recognize that our data are incomplete (e.g. Neiber contain all the possible oceanic archipelagos), or alternatively et al., 2011). However, insofar as the data are incomplete, or that one is not interested in generalizing the results to other subject to excessive taxonomic splitting or lumping, we levels of the factor not included in the study (e.g. other oceanic assume that there is unlikely to be significant bias across the archipelagos not included in the study). When a factor is stud- islands within a particular archipelago as the same taxono- ied as a random effect, instead of fitting regression parameters mists usually work on material from all the islands within an for each level of a factor, one is interested in estimating the archipelago. Small numbers of additional or deleted species variation in the regression parameters induced by the different make little difference to the trends shown here. A possible levels of the factor (e.g. the variation around the general slope exception is shown by Cowie’s (1995) path analysis of varia- considering all the islands belonging to the same virtual global tion in species richness of Hawaiian land snails, in which the archipelago). The random effects can be seen as grouping fac- densely populated island of Oahu appears oversampled in tors drawn as a random sample from a larger (conceptual) comparison to the much larger but sparsely populated island population, such as the eight archipelagos considered here of Hawaii. (adapted from glossary in Bunnefeld & Phillimore, 2012). In this context, one is not primarily interested in estimating and comparing the relationships under study for the different levels Island groupings of the factor. The random effect is seen as a source of pseu- The configurations of islands vary through time, not just doreplication [non-independence of data points (here islands) because of the ontogeny of the islands, but also due to the belonging to the same level of the factor (here archipelagos)] influence of other factors, e.g. tectonics and eustatic change. that needs to be taken into account. In particular, sea-level minima during the Pleistocene pro- Bunnefeld & Phillimore (2012) demonstrated the advanta- duced connections between some adjacent islands, turning ges of LMMs when applied to the data that were originally 2 them into single islands, while volcanism can both join and used for testing the ATT model (i.e. D = b1 + b2Area + 2 sometimes subdivide island territories (general review in b3Time + b4Time ). Here we follow the same methodological Whittaker & Fernańdez-Palacios, 2007). For some analyses steps: our response variables were the three diversity metrics (detailed below), we have treated such groups as single (SR, nEnd and nSIE). The fixed effects were island area islands so as to test the possible effects of varying configura- (Area, in km2; log-transformed as generally supported by tion of the archipelagos through time (see Appendix S1, previous work, e.g. Whittaker et al., 2008; Triantis et al., Table S2 & Appendix S2). 2012a), time elapsed since island formation (Time, i.e. date of emergence of each island, in million years ago, Ma), a quadratic term for the time elapsed (Time2), and we also Statistical analyses considered elevation (Elevation, in metres, m) as a proxy of In studies such as this, the small number of islands per archi- environmental heterogeneity. The grouping factor considered pelago can lead to low power in detecting trends, instability in as a random effect was the archipelago that each island parameter estimation and model over-fitting (e.g. Burnham & belongs to, as the values of the intercept and the slopes of Anderson, 2002). Bunnefeld & Phillimore (2012) recently sug- the relationships between the diversity metrics, area, time gested the use of linear mixed effect models (LMMs) to over- and elevation may vary across archipelagos. To select the best come such limitations. LMMs are designed to detect general models for describing the diversity metrics we followed a patterns where data come from grouped sources (Bolker et al., two-step procedure: first, the most parsimonious random 2009; Zuur et al., 2009). We follow their lead in applying effects structures (with all fixed effects included) were LMMs to allow the simultaneous consideration of data from selected using model selection based on the small-sample eight archipelagos comprising 56 islands. corrected Akaike’s information criterion (AICc) (Burnham &

LMM predictors are classified into (1) fixed effects: those Anderson, 2002). The model with the lowest AICc value is for which we aim to estimate regression parameters, i.e. slope considered to fit the data best. However, all models with a and intercept; and (2) random effects: those that identify ΔAICc value < 2 (the difference between each model’s AICc groups conceptually drawn from a larger population (e.g. and the lowest AICc) must be considered as having relatively archipelagos or taxa) within the data and for which we exam- similar levels of support and thus belong to the group of ‘best ine variation in a parameter (i.e. slope and intercept) across models’ (i.e. equally parsimonious; Burnham & Anderson, levels (Bunnefeld & Phillimore, 2012; Hortal, 2012). When 2002). Accordingly, when several random structures provided considering a factor as fixed we are interested in estimating indistinguishable AICc values (ΔAICc value < 2), they were

120 Journal of Biogeography 40, 117–130 ª 2012 Blackwell Publishing Ltd Land snails on oceanic islands considered as part of the ‘best random structure group’, and the diversity metric considered) and also for the untrans- subsequent fixed effect structures were compared. To find formed values. As we employ log-area in the analysis, this the most parsimonious random effect structures we com- particular implementation of the ATT² assumes a power law pared models with and without a varying intercept among species–area relationship, the most general and widely archipelagos and all possible combinations of varying slopes applied species–area relationship model (Rosenzweig, 1995; across archipelagos for the different variables considered (i.e. Triantis et al., 2012a). log(Area), Time, Time² and Elevation). We used the ‘lmer’ We have undertaken additional analyses for separate snail function in the ‘lme4’ library (version 0.999375-39) in R families for the two most species-rich archipelagos, i.e. 2.14.1 (R Development Core Team, 2011) using restricted Hawaii (10 islands) and the Canaries (7 islands), following maximum likelihood (REML). the above steps, but with the random effect being ‘Family’ After determining the best random effect structures, the instead of ‘Archipelago’, in order to compare the diversity most parsimonious combinations of fixed effects were found patterns within the same archipelago but for different taxo- using model selection based on AICc (models fitted using max- nomic groupings. For the Hawaiian Islands the taxonomic imum likelihood). We used the ‘dredge’ function in the groupings considered were: Succineidae, Pupilloidea, Helicar- ‘MuMIn’ library in R (version 0.13.17) to run a complete set ionidae, Helicinidae, Endodontoidea, Amastridae and Achati- of models with all possible combinations of the fixed effects nellidae. For the Canary Islands they were: Vitrinidae, and determined the subset of ‘best models’ as the ones with Helicoidea, Ferussaciidae and Enidae.

ΔAICc value < 2 (as above). We additionally used Akaike weights derived from the AIC (wAIC ) to evaluate the relative c c RESULTS likelihood of each model, given the dataset and the set of models considered, and to estimate the relative importance of each Linear mixed effect models: archipelagos variable by summing these wAICc across the models in which they were included. Akaike weights are directly interpreted in Random effects terms of each model’s probability of being the best at explain- ing the data (Burnham & Anderson, 2002). For the log-transformed values of the diversity metrics, the

To facilitate comparison with the previous LMM applica- random effect structures selected based on the lowest AICc tions of the ATT2 (Bunnefeld & Phillimore, 2012), we have scores were as follows. For SR two random effect structures applied the above methodological steps for the log-trans- were selected, a random intercept and a random slope for formed values of the diversity metrics considered (log n + 1 log(Area), and a random slope for log(Area), separately. For was used for datasets that contain at least one zero value for nSIE only a random slope for log(Area) was selected. For

Table 2 Parameter estimates for the ﬁxed effects of the most parsimonious linear mixed effect models for the land snails of the 56 islands considered (eight archipelagos). The models with a DAICc of less than two are presented (see Materials and Methods). The random structure (intercept or slopes varying across archipelagos), the number of parameters in the model (p), AICc and the AICc difference (DAICc) and Akaike weights (wAICc) are given for each model. NI indicates that the variable was not included in the model. The results remain identical for island groupings (see Appendix S3). The ATT2 model is highlighted in bold. All diversity metrics were log-transformed. The importance of each variable is estimated by summing, for each combination of diversity metric and random structure, the Akaike weights of the models in which it was included. Time refers to the maximum surface age of each system (see Table 1).

2 Metric Random structure Intercept Time Time Log(Area) Elevation p AICc DAICc wAICc

1. SR Log(Area) 0.414 0.039 À0.003 0.471 <0.000 7 22.131 0.000 0.369 0.359 0.051 À0.003 0.424 NI 6 22.185 0.054 0.359 0.538 NI À0.001 0.481 <0.000 6 22.736 0.605 0.272 Variable importance 0.727 1 1 0.641 2. SR Intercept & Log(Area) 0.491 0.054 À0.003 0.357 NI 8 22.383 0.000 0.449 0.537 0.042 À0.003 0.403 <0.000 9 22.719 0.336 0.380 0.672 NI À0.001 0.417 <0.000 8 24.328 1.945 0.170 Variable importance 0.829 1 1 0.550 1. nEnd Log(Area) 0.322 0.060 À0.003 0.393 NI 6 26.049 0.000 1 Variable importance 1 1 1 0 2. nEnd Intercept 0.279 0.055 À0.003 0.431 NI 6 26.649 0.000 0.71 0.323 0.047 À0.003 0.462 <0.000 7 28.439 1.789 0.29 Variable importance 1 1 1 0.290 1. nSIE Log(Area) À0.150 0.109 À0.005 0.359 NI 6 43.034 0.000 1 Variable importance 1 1 1 0

SR, number of native species; nEnd, number of archipelagic endemic species; nSIE, number of single-island endemic species; Area, island area in km2.

Journal of Biogeography 40, 117–130 121 ª 2012 Blackwell Publishing Ltd R. A. D. Cameron et al. nEnd, two random effect structures were selected: a random Interestingly, for the log-transformed values of the diversity intercept and a random slope for log(Area), separately metrics, no simpler model than the ATT2 (such as species– (Table 2). The random effects variation distribution for the area or species–area–time) was included in the group of the linear mixed effect models selected is given in Table S3 in best models. This is the only major difference between the Appendix S3. The results are similar for the island groupings results for transformed and untransformed diversity values (results not shown). because in some cases simpler models were selected for the untransformed diversity metrics (results not shown). The results are similar when all islands are considered individu- Fixed effects ally and when the island groupings are considered (Table S4 2 According to the AICc-based model selection procedure, the in Appendix S3). Considering the coefﬁcients for the ATT ATT2 model was always amongst the most parsimonious for model for the three different diversity metrics, using the all the diversity metrics considered (Table 2; see also Fig. 2). same random structure, i.e. random slope for log(Area), we

Figure 2 Species–area–time relationships (ATT2) for the native land snail species of the eight oceanic archipelagos considered in this study. The surface is the prediction from the linear mixed effect model including a random slope for log(Area) varying across archipelagos (Table 2), with number of species and area log-transformed. Points represent the observed data. Time is date of emergence of each island in million years ago (Ma). For panels (a)–(h) the colour shading indicates the species richness predicted by the model [from white (low) to red (high)]. Note that for panel (i) the slope corresponds to the grand mean ﬁtted by the linear mixed effect model.

122 Journal of Biogeography 40, 117–130 ª 2012 Blackwell Publishing Ltd Land snails on oceanic islands

Table 3 Parameter estimates for the ﬁxed effects of the most parsimonious linear mixed effect models based on AICc values for Hawaii (10 islands and seven families) and the Canary snail families (seven islands and four families; see text). The random structure (intercept or slopes varying across families), the number of parameters in the model (p), and the AICc difference (DAICc) and Akaike weights (wAICc) are given for each model. NI indicates that the variable was not included in the model. The results based on the untransformed diversity values remain identical. Diversity metrics were log-transformed. A variable importance is estimated by summing, for each combination of diversity metric and random structure, the Akaike weights of the models in which it was included.

Random 2 Archipelago structure Intercept Time Time Elevation Log(Area) p AICc DAICc wAICc

Hawaii Log(Area) À0.600 0.134 À0.005 NI 0.426 11 66.595 0.000 1 Variable importance 1 1 0 1 Canaries Intercept 0.476 0.083 À0.004 NI NI 5 35.912 0.000 0.435 0.710 NI NI NI NI 3 36.060 0.149 0.403 0.586 NI NI <0.000 NI 4 37.900 0.161 0.160 Variable importance 0.435 0.435 0.160 0

Time, geological age of the island in million years since emergence; Area, island area in km2. observe: (1) a progressive decrease of the log(Area) effect in groups of organisms due to factors other than time and area turn considering natives, archipelagic endemics and SIE, and (and aspects of environment and history closely correlated (2) an increase of the effect of Time2, and especially of Time, with them) that may inﬂuence the diversity and composition from natives to SIE. of the biotas. The theory does not yet allow us to generate Among the selected models, log(Area) and Time² were precise predictive models in the absence of relevant, quanti- always included and consequently have maximum relative tative data on properties of the taxa considered (cf. Rosindell variable importance values. Time and elevation were always & Phillimore, 2011). Thus, differences in the responses of the less important variables, with elevation having lower biotas to time and area among both archipelagos and taxa importance across all models (Table 2). not only interrogate the model, but also offer clues to under- standing variation in the relative magnitude of the inﬂuence of the factors involved. Here, we examine the patterns shown Linear mixed effect models: taxonomic subdivisions in relation to the GDM and then consider the outstanding issues that arise from our analyses. Random effects For the log-transformed values of the species richness for the Snails, oceanic archipelagos and the GDM main taxonomic groups for Hawaii and the Canaries the random effect structures selected were: Hawaii, random slope Within the framework of the GDM, the shape of the rela- for log(Area); the Canaries, random intercept (Table 3). tionship between key diversity metrics and time can vary according to the degree to which the archipelago encom- passes the full range of island stages (Fig. 1; see: Whittaker Fixed effects et al., 2008; Borges & Hortal, 2009; Fattorini, 2009; Triantis

According to the AICc-based model selection procedure, for et al., 2010). The two models (Fig. 1a, c) considering only a Hawaii the most parsimonious model was the ATT2 linear relationship with time (either positive or negative) can (Table 3, see Fig. 3). By contrast, TT2 was the most parsimo- be viewed as special cases of the general model (Fig. 1b). For nious model for the Canarian taxonomic groupings (Table 3, the data considered here, the ATT2 model has in most cases Fig. 4), signifying area as a non-essential descriptor of snail greater generality than the traditional richness–area models, richness on this archipelago. For untransformed values, sim- or time-only models. Plotting the ATT2 for the native species pler models than ATT2 were sometimes within the group of richness (with a random slope for log(Area) only; Table 2) best fit, although the ATT2 was always amongst the most for all the archipelagos (Fig. 2) illustrates the positive rela- parsimonious models (results not shown). tionship with area, and the full range of relationships with time, from hump-shaped to almost linear (as Fig. 1). While the ATT2 formulation thus provides a good general descrip- DISCUSSION tion of the data, it should be emphasized that we cannot rule The general dynamic model of oceanic island biogeography out alternative models (cf. Bunnefeld & Phillimore, 2012). seeks to provide a robust framework within which to con- All the archipelagos considered are composed of true oce- sider species diversity of oceanic island biotas but at the anic islands, yet their mechanisms of formation vary, con- same time it is an intentionally simplified representation of tributing to inter-archipelago differences in diversity both the geological development and of species diversity processes and patterns. For all the diversity metrics, and for dynamics (Whittaker et al., 2008, 2010). Details of pattern all the archipelagos and island groupings, the most parsimo- are expected to vary both among archipelagos and among nious random effect structures included invariant coefficients

Journal of Biogeography 40, 117–130 123 ª 2012 Blackwell Publishing Ltd R. A. D. Cameron et al.

Figure 3 Species–area–time relationships (ATT2) for the richness of the native species of the seven land snail families of the Hawaiian Islands: (a) for all the 10 islands considered, and (b) only for the six large islands. For the former the surface is the prediction from the linear mixed effect model including a random slope for log(Area) varying across families (Table 3). For the latter, the surfaces are the prediction from a model including a random slope for Elevation, Time and Time2 varying across families and including log(Area), Elevation, Time and Time2 as ﬁxed effects: elevation values were averaged and multiplied by the respective coefﬁcients for each family to produce the surfaces. Points represent the observed data. The numbers of species and area values are log-transformed. Time is in million years ago (Ma). The colour shading indicates the species richness predicted by the model [from white (low) to red (high)].

124 Journal of Biogeography 40, 117–130 ª 2012 Blackwell Publishing Ltd Land snails on oceanic islands

Figure 4 Species–time relationships (TT2) for the richness of the native species of the four land snail families of the Canary Islands. The solid line represents the most parsimonious model based on the data for all families (see Table 3). Time is date of emergence of each island in million years ago, Ma (see Materials and Methods for details on island groupings). The number of species (S) is log- transformed. for both Time and Time² across archipelagos (Table 2). This Taking two islands of the same size, Madeira and Saõ indicates the consistency of the effect of geological age on Miguel (Azores), Cameron et al. (2007) has previously the processes generating and maintaining diversity of snails, accounted for differences in snail richness as a product of regardless of the particularities of the geological setting in effective island age and the range of available habitats. each archipelago considered here. Especially for the archipe- While the notional geological age of both islands is c. lagic endemic species and the SIE, this result signifies the 4.5 Ma, Saõ Miguel reached its current shape just 0.05 Ma, importance of island areas to evolutionary dynamics (e.g. by the formation of a land bridge between an older eastern Rosenzweig, 1995; Triantis et al., 2008a), and at the same island, only part of which originated 4 Ma and a younger time suggests that the inter-archipelagic differences in abso- western island that originated 0.55 Ma. It has even been lute geological ages are less important. On the other hand suggested that the whole island is less than 1 Myr old the sequential increase of the time effect as we move from (Johnson et al., 1998). Given all of which, the expectation natives to SIE (Table 2) signifies the critical role of time for would be that Saõ Miguel is still accumulating diversity. the formation of new species. This is indeed confirmed by detailed anatomical and molec- For Hawaii, Canaries, Madeira and Tristan da Cunha, ular studies within generic clades (Martins, 2005; Van Riel diversity rose and fell with increasing time, as predicted by et al., 2005; Jordaens et al., 2009). The relationship of age the GDM (Table 2, Fig. 2). For the other archipelagos, i.e. and ontogeny within the Azores in fact demands further Azores, Gala´pagos, Samoa and Society, we found only an comment. Assuming Saõ Miguel to be in part around increase with time (Fig. 2; see also Borges & Hortal, 2009; 4 Myr old, the only other old island is Santa Maria, which Triantis et al., 2010; Bunnefeld & Phillimore, 2012). The is estimated to have a maximum age of 8.12 Ma (Borges main dichotomy between the two main groups of archipela- et al., 2009). However, extensive volcanic activity occurred gos identified above is the presence or absence of a full range on Santa Maria around 4 Ma. Thus, the entire Azorean of island stages, from ‘young’ to ‘old’ islands, which is not archipelago is essentially still in relative youthful stages of simply a matter of the time elapsed since island origination. the island life cycle, while more than 60% of the land area For example Terceira in the Azores (c. 3.5 Ma, 400 km2)isa of the archipelago is < 1 Myr old. Nevertheless, a synthetic volcanically active island, whereas Bora Bora, an island of consideration of these and other explanations (e.g. climatic similar age in the Society Islands (c. 3.4 Ma, 30 km2), is a effects) are required for a full explanation of these patterns highly eroded, inactive shield volcano. (e.g. Triantis et al., 2012b).

Journal of Biogeography 40, 117–130 125 ª 2012 Blackwell Publishing Ltd R. A. D. Cameron et al.

the complexity of the geological and ecological processes The effects of taxonomic subdivision involved and the diverse histories of colonization and specia- Plotting the ATT2 for the native species richness for the seven tion and extinction, there are a number of exceptions. The taxonomic groups in Hawaii illustrates a consistent pattern of challenge is thus to identify the factors involved in oceanic positive relationships with area and hump-shaped relation- island diversity dynamics without resorting to a mass of ad ships with time (Fig. 3a). The pattern owes much to the diver- hoc speculations and post hoc rationalizations. Such factors sity differences between the large main islands of the will include aspects of the islands themselves, and, as shown archipelago and the small north-west islands. When the small already for the Azores (Borges & Hortal, 2009; Triantis et al., islands are removed, the relationship with time varies from 2010), differences in life histories and dispersal abilities highly humped for Amastridae and Achatinellidae, almost lin- among subsets of organisms. We see the following as themes ear for Succinidae, to a shallow u-shape for Helicinidae worth further consideration in this context. (Fig. 3b). The overall humped pattern with island age is thus 1. The GDM is cast at the level of the emergent whole-sys- demonstrated to be due largely to the two richest families, the tem dynamics, and disaggregation into taxonomically or eco- Achatinellidae and Amastridae (Table S1 in Appendix S1). logically distinct groups may be expected to show varying The former, tree snails confined to the Pacific Basin, and the responses to long-term island-environmental dynamics (e.g. latter, all endemic to Hawaii, appear to have very limited pow- Borges & Hortal, 2009). Ideally, hypotheses for such distinc- ers of dispersal. Succineids tend to occupy open and at least tive responses should be formulated in advance of analysis seasonally wet habitats: they are most diverse on the youngest based on ecological characteristics of the groups concerned. (and largest) island where these habitats are widely available. 2. Analyses should seek to place alternative hypotheses into The ecology of the Helicinidae, which fits poorly with the a multiple working hypothesis framework. For instance, Kim ATT2 model (Fig. 3b) is little known (see Cowie, 1996). et al. (2008) identified three discrete waves of colonization For the Canaries, the general success of the TT2 model by monophyletic endemic plant lineages of Macaronesian (see Results) again demonstrates some interesting variability islands from the western Mediterranean, offering support to according to family, with the Ferussaciidae standing out as the ‘colonization window hypothesis’ (Carine, 2005), accord- an exception for having their greatest richness on the oldest ing to which, the opportunity for island colonization may islands (Fuerteventura and Lanzarote; Fig. 4). The Ferussacii- have been largely constrained to one or more distinct periods dae are usually found in subterranean environments in warm of time (e.g. glacial era sea-level low stands; Fernańdez-Pala- and dry areas, more typical of the two oldest islands, in part cios et al., 2011). This alternative theoretical case to the reflecting their topography and in part their more easterly GDM makes specific predictions regarding the temporal location. In contrast, forest-dwelling forms such as vitrinids development of island phylogenies distinct from the GDM, are now entirely missing from Fuerteventura and Lanzarote which should be testable if genetic analyses of multiple sepa- and are more diverse in the intermediate-aged forested rate clades are undertaken (Whittaker et al., 2010). islands such as Tenerife. Similar patterns are observed in the 3. Other variables (apart from area) can be used for a more Madeiran archipelago, with the composition of the Porto effective approximation of the available ecological space, e.g. Santo and Desertas faunas resembling those of the dry east- habitat diversity. For example, in a recent study of taxa from ern islands of the Canaries. Moreover, while Porto Santo is a variety of island groups (including the Canaries, Azores much smaller in size than the most humid island of the and Cape Verde islands), Triantis et al. (2008b, 2010) con- group, i.e. Madeira, the dominant superfamily, Helicoidea, is cluded that the precise quantification of factors such as cli- actually richer on Porto Santo (see Cook, 2008). mate, habitat diversity and evolutionary history (that may Interestingly, the taxonomic decomposition of the snail fau- partially covary with area) is necessary to develop a more nas of Hawaii and the Canaries revealed that the coefficients predictive model of how species numbers vary across insular associated with time vary across the different taxonomic systems (see also Losos & Parent, 2010). groupings (see Table 3 and note that for the Canaries there is 4. According to the GDM, lineage radiation (leading to not a single best random structure, although one of them multiple SIEs on individual islands) should be most preva- includes time). This is in contrast to the structure selected for lent after the initial colonization phase, in the period leading the snail faunas across the different archipelagos. These find- up to island maturity, coinciding with maximal species carry- ings indicate that whatever the emergent whole system pat- ing capacity and the development of maximal topographic tern, within the same archipelago the relationship of the complexity (see also Fig. 4.5 in Whittaker et al., 2010). There diversity of the different taxonomic groupings with time var- is support to be found for this general scenario (e.g. see Giv- ies in accordance with their different ecological requirements. nish et al., 2009), although the specific characteristics of each island and archipelago play a significant role in the timing and the magnitude of radiation events. Whittaker et al. Future directions and hypotheses (2008, p. 983) argued that the GDM predicts that ‘Adaptive In general, we have shown that the ATT2 formalization of radiation (AR) will be the dominant process on islands the GDM can successfully describe patterns in snail diversity where the maximum elevational range occurs, as it generates among islands within these archipelagos. Inevitably, given the greatest richness of habitats (major ecosystem types),

126 Journal of Biogeography 40, 117–130 ª 2012 Blackwell Publishing Ltd Land snails on oceanic islands including novel ones that few colonists have experienced, (Gastropoda, Pulmonata, Enidae). Biological Journal of the whereas non-adaptive radiation (NAR) will become relatively Linnean Society, 89, 169–187. more important on slightly older islands, past their peak ele- Anderson, C.L., Channing, A. & Zamuner, A.B. (2009) Life, vation, owing to increased topographical complexity promot- death and fossilization on Gran Canaria – implications for ing intra-island allopatry’. Testing these predictions provides Macaronesian biogeography and molecular dating. Journal a significant challenge (but see Givnish et al., 2009; Rabosky of Biogeography, 36, 2189–2201. & Glor, 2010). Bolker, B.M., Brooks, M.E., Clark, C.J., Geange, S.W., Poul- 5. There is some evidence of variation in the effectiveness sen, J.R., Stevens, H.H. & White, J.-S.S. (2009) Generalized and dynamism of habitat barriers within islands in different linear mixed models: a practical guide for ecology and archipelagic settings. For instance, in the Gala´pagos, the first evolution. Trends in Ecology and Evolution, 24, 127–135. stages of adaptive differentiation within bulimulid species can Borges, P.A.V. & Hortal, J. (2009) Time, area and isolation: be detected in the early history of islands, related to the avail- factors driving the diversification of Azorean arthropods. ability of as yet unoccupied niches (Parent & Crespi, 2009). Journal of Biogeography, 36, 178–191. Evidence from other archipelagos, however, suggests that Borges, P.A.V., Amorim, I.R., Cunha, R., Gabriel, R., Mar- non-adaptive radiation in snails can also peak at an early tins, A.M.F., Silva, L., Costa, A. & Vieira, V. (2009) Azores stage. For example, Martins (2005) suggested that recurrent – biology. Encyclopedia of islands (ed. by R. Gillespie and volcanism played a significant part in the build-up of single- D.A. Clague), pp. 70–75. University of California Press, island diversity in the relatively young Azorean snail faunas. Berkeley, CA. There may be several episodes of partial differentiation fol- Bunnefeld, N. & Phillimore, A.B. (2012) Island, archipelago lowed by some introgression in early stages. Later, the rate of and taxon effects: mixed models as a means of dealing differentiation/speciation slows down (Martins, 2011). Cook with the imperfect design of nature’s experiments. Ecogra- (2008) similarly suggested that a coincidence of the rates of phy, 35,15–22. volcanic and erosive events with the speed of genetic differen- Burnham, K.P. & Anderson, D.R. (2002) Model selection and tiation in snail populations (the ‘geodetic rate’) accounts for multimodel inference: a practical information-theoretic diversity in the Madeiran snail fauna (see also Carson et al., approach, 2nd edn. Springer-Verlag, New York. 1990, on drosophilids in Hawaii). There is some evidence in Cameron, R.A.D., Cook, L.M. & Hallows, J.D. (1996) Land these studies of initially high rates of diversification, facili- snails on Porto Santo: adaptive and non-adaptive radia- tated by temporary barriers caused by intermittent volcanic tion. Philosophical Transactions of the Royal Society B: Bio- episodes, and by the exceptionally poor powers of active dis- logical Sciences, 351, 309–327. persal in snails (Kisel & Barraclough, 2010). Once again, the Cameron, R.A.D., da Cunha, R.M.T. & Frias Martins, A.M. challenge will be to develop models that retain generality in (2007) Chance and necessity: land-snail faunas of Saõ the face of such system-specific differences. Miguel, Azores, compared with those of Madeira. Journal of Molluscan Studies, 73,11–21. Cardoso, P., Arnedo, M., Triantis, K.A. & Borges, P.A.V. ACKNOWLEDGEMENTS (2010) Drivers of diversity in Macaronesian spiders and We are grateful to Stan Hart, Jose´ Marıá Fernańdez-Palacios, the role of species extinctions. Journal of Biogeography, 37, Dennis Geist and Richard Preece for access to unpublished 1034–1046. work and advice on palaeogeography and/or malacofauna of Carine, M.A. (2005) Spatio-temporal relationships of the particular archipelagos. Special thanks to Nils Bunnefeld for Macaronesian endemic flora: a relictual series or window helping us with the application of LMMs and Ally Phillip- of opportunity? Taxon, 54, 895–903. more, Michal Horsa´k, Menno Schilthuizen and an anony- Carracedo, J.C., Pe´rez Torrado, F.J., Ancochea, E., Meco, J., mous referee for comments and discussion on a previous Hernań, F., Cubas, C.R., Casillas, R., Rodrı´guez Badiola, E. version of the manuscript. K.A.T. was supported in this work & Ahijado, A. (2002) Cenozoic volcanism II: the Canary by a Marie Curie Intra-European Fellowship Program (pro- Islands. The geology of Spain (ed. by W. Gibbons and T. ject ‘SPAR’, 041095) held in the School of Geography and Moreno), pp. 439–472. The Geological Society, London. the Environment, University of Oxford, and by a Fundaçaõ Carson, H.L., Lockwood, J.P. & Craddock, E.M. (1990) para a Cieˆncia e a Tecnologia (FCT) Fellowship (SFRH/BPD/ Extinction and recolonization of local populations on a 44306/2008). growing shield volcano. Proceedings of the National Acad- emy of Sciences USA, 87, 7055–7077. Chiba, S. (1999) Accelerated evolution of land snails Manda- REFERENCES rina in the oceanic Bonin Islands: evidence from mito- – Alonso, M.R., Goodacre, S.L., Emerson, B.C., Ibań˜ez, M., chondrial DNA sequences. Evolution, 53, 460 471. Hutterer, R. & Groh, K. (2006) Canarian land snail diver- Clague, D.A. (1996) The growth and subsidence of the sity: conflict between anatomical and molecular data on Hawaiian-Emperor volcanic chain. The origin and evolution the phylogenetic placement of five new species of Napaeus of Pacific island biotas, New Guinea to Eastern Polynesia:

Journal of Biogeography 40, 117–130 127 ª 2012 Blackwell Publishing Ltd R. A. D. Cameron et al.

patterns and processes (ed. by A. Keast and S.E. Miller), R.J. (2011) A reconstruction of Palaeo-Macaronesia, with pp. 35–50. SPB Academic Publishing BV, Amsterdam. particular reference to the long-term biogeography of the Clouard, V. & Bonneville, A. (2005) Ages of seamounts, Atlantic island laurel forests. Journal of Biogeography, 38, islands and plateaus on the Pacific Plate. Plates, plumes 226–246. and paradigms (ed. by G.R. Foulger, J.H. Natland, D. Gargominy, O. (2008) Beyond the alien invasion: a recently Presnall and D.L. Anderson), pp. 71–90. Geological Society discovered radiation of Nesopupinae (Gastropoda: Pulmo- of America Special Paper, 388. Boulder, CO. nata: Vertiginidae) from the summits of Tahiti (Society Cook, L.M. (2008) Species richness in Madeiran land snails, Islands, French Polynesia). Journal of Conchology, 39, 517– and its causes. Journal of Biogeography, 35, 647–653. 536. Coote, T. & Loe`ve, E. (2003) From 61 species to five: ende- Geist, D. (1996) On the emergence and submergence of the mic tree snails of the Society Islands, French Polynesia fall Galapagos Islands. Noticias de Galapagos, 56,5–8. prey to an ill-judged biological control programme. Oryx, Geldmacher, J., Hoernle, K., Bogaard, P.V.D., Duggen, S. & 37,91–96. Werner, R. (2005) New 40Ar/39Ar age and geochemical Coppois, G. (1985) E´tude de la spećiation chez les Bulimulidae data from seamounts in the Canary and Madeira volcanic ende´miques de l’archipel des Galapagos (Mollusques, province: support for the mantle plume hypothesis. Earth Gaste´ropodes, Pulmone´s). Faculte´ des Sciences, Universite´ and Planetary Science Letters, 237,85–101. Libre, Bruxelles. Givnish, T.J., Millam, K.C., Mast, A.R., Paterson, T.B., Courtillot, V., Davaille, A., Besse, J. & Stock, J. (2003) Three Themi, T.K., Hipp, A.L., Henss, J.M., Smith, J.F., Wood, distinct types of hotspots in the Earth’s mantle. Earth and K.R. & Sytsma, K.J. (2009) Origin, adaptive radiation and Planetary Science Letters, 206, 295–308. diversification of the Hawaiian lobeliads (Asterales: Cam- Cowie, R.H. (1995) Variation in species diversity and shell panulaceae). Proceedings of the Royal Society B: Biological shape in Hawaiian land snails: in situ speciation and eco- Sciences, 276, 407–416. logical relationships. Evolution, 49, 1191–1202. Goodfriend, G.A., Cameron, R.A.D., Cook, L.M., Courty, Cowie, R.H. (1996) Pacific island land snails: relationships, M.-A., Federoff, N., Kaufman, N., Livett, E. & Tallis, J. origins and determinants of diversity. The origin and evolu- (1996) The Quaternary eolianite sequence of Madeira: tion of Pacific island biotas, New Guinea to eastern Polynesia: stratigraphy, chronology, and paleoenvironmental interpre- patterns and processes (ed. by A. Keast and S.E. Miller), pp. tation. Palaeogeography, Palaeoclimatology, Palaeoecology, 347–372. SPB Academic Publishing, Amsterdam. 120, 195–234. Cowie, R.H. (1998) Catalog of the non-marine snails and Holdgate, M.W. (1965) The fauna of the Tristan da Cunha slugs of the Samoan Islands. Bishop Museum Bulletin in islands. Philosophical Transactions of the Royal Society of Zoology, 3,1–122. London Series B, Biological Sciences, 249, 361–402. Cowie, R.H. (2001) Decline and homogenization of Pacific Holyoak, D.T. & Holyoak, G.A. (2009) A new species of Lau- faunas: the land snails of American Samoa. Biological Con- ria (Gastropoda, Lauriidae) from the Canary Islands. Ibe- servation, 99, 207–222. rus, 27,1–5. Cowie, R.H., Evenhuis, N.L. & Christensen, C.C. (1995) Cat- Hortal, J. (2012) Escaping the trap of low sample size in alog of the native land and freshwater molluscs of the island biogeography. Frontiers of Biogeography, 3, 127–129. Hawaiian Islands. Backhuys, Leiden. Johnson, C.L., Wijbrans, J.R., Constable, C.G., Gee, J., Stau- Cowie, R.H., Rundell, R.J., Mika, F. & Setu, P. (2002) The digel, H., Tauxe, L., Forjaz, V.H. & Salgueiro, M. (1998) endangered partulid tree snail Samoana thurstoni on Olos- 40Ar/39Ar ages and paleomagnetism of Saõ Miguel lavas, ega and the land snail diversity of the Manu’a Islands, Azores. Earth and Planetary Science Letters, 160, 637–649. American Samoa. American Malacological Bulletin, 17,37– Jordaens, K., Van Riel, P., Martins, A.M.F. & Backeljau, T. 43. (2009) Speciation on the Azores islands: congruent pat- Cunha, R., Rodrigues, A. & Martins, A.M.F. (2010) List of terns in shell morphology, genital anatomy, and molecular Molluscs. A list of the terrestrial and marine biota from the markers in endemic land snails (Gastropoda, Leptaxinae). Azores (ed. by P.A.V. Borges, A. Costa, R. Cunha, R. Gab- Biological Journal of the Linnean Society, 97, 166–176. riel, V. Gonçalves, A.F. Martins, I. Melo, M. Parente, P. Kim, S.-C., McGowen, M.R., Lubinsky, P., Barber, J.C., Raposeiro, P. Rodrigues, R.S. Santos, L. Silva, P. Viera and Mort, M.E. & Santos-Guerra, A. (2008) Timing and tempo V. Viera), pp. 165–177. Princı´pia, Oeiras. of early and successive adaptive radiations in Macaronesia. Dall, W.H. & Ochsner, W.H. (1928) Landshells of the Gala- PLoS ONE, 3, e2139. doi:10.1371/journal.pone.0002139. pagos Islands. Proceedings of the California Academy of Sci- Kisel, Y. & Barraclough, T.G. (2010) Speciation has a spatial ences, 17, 141–185. scale that depends on levels of gene flow. The American Fattorini, S. (2009) On the general dynamic model of oce- Naturalist, 175, 316–334. anic island biogeography. Journal of Biogeography, 36, 1100 Losos, J.B. & Parent, C.E. (2010) The speciation–area rela- –1110. tionship. The theory of island biogeography revisited (ed. by Fernańdez-Palacios, J.M., de Nascimento, L., Otto, R., Del- J.B. Losos and R.E. Ricklefs), pp. 415–438. Princeton Uni- gado, J.D., Garcıá-del-Rey, E., Are´valo, J.R. & Whittaker, versity Press, Princeton, NJ.

128 Journal of Biogeography 40, 117–130 ª 2012 Blackwell Publishing Ltd Land snails on oceanic islands

MacArthur, R.H. & Wilson, E.O. (1967) The theory of island Azores islands. Molecular Phylogenetics and Evolution, 37, biogeography. Princeton University Press, Princeton, NJ. 132–143. Marrero, A. & Francisco-Ortega, J. (2001) Evolucioń en islas: Rosenzweig, M.L. (1995) Species diversity in space and time. la forma en el tiempo. Naturaleza de las islas Canarias: Cambridge University Press, Cambridge, UK. ecologıá y conservacioń (ed. by J.M. Fernańdez-Palacios and Rosindell, J. & Phillimore, A.B. (2011) A unified model of J.L. Martıń Esquivel), pp. 141–150. Turquesa Ediciones, island biogeography sheds light on the zone of radiation. Santa Cruz de Tenerife. Ecology Letters, 14, 552–560. Martins, A.M.F. (2005) The shaping of a species: the Azorean Ryan, P.G. (2009) Tristan da Cunha and Gough Island. Drouetia Gude (Pulmonata: Zonitidae: Oxychilus)asa Encyclopedia of islands (ed. by R.G. Gillespie and D.A. Cla- model. Records of the Western Australian Museum, Supple- gue), pp. 922–932. University of California Press, Berkeley, ment, 68, 143–157. CA. Martins, A.M.F. (2011) When the Gala´pagos “finches” are Seddon, M.B. (2008) The landsnails of Madeira: an illustrated Azorean snails. Açoreana, Suplemento, 7, 209–228. compendium of the landsnails and slugs of the Madeiran Neall, V.E. & Trewick, S.A. (2008) The age and origin of the archipelago. Studies in Biodiversity and Systematics of Ter- Pacific islands: a geological overview. Philosophical Trans- restrial Organisms from the National Museum of Wales actions of the Royal Society B: Biological Sciences, 363, 3293 (NMW), BIOTIR Report 2. NMW, Cardiff. –3308. Smith, A.G. (1966) Land snails of the Galapagos. The Neiber, M.T., Vega-Luz, R., Vega-Luz, R. & Koenemann, S. Galapagos. Proceedings of the Symposia of the Galapagos (2011) Hemicycla (Adiverticula) diegoi (Gastropoda: Pul- International Scientific Project, 1964 (ed. by R.I. Bow- monata: Helicidae), a new species from Tenerife, Canary man), pp. 240–251. University of California Press, Islands, with a phylogenetic analysis of conchologically Berkeley, CA. similar species in the genus Hemicycla Swainson, 1840. Solem, A. (1984) A world model of land snail diversity and Zootaxa, 2757,29–46. abundance. World-wide snails: biogeographical studies on Nu´ n˜ez, L. & Nu´ n˜ez, J. (2010) Mollusca. Lista de especies sil- non-marine molluscs (ed. by A. Solem and A.C. van Brug- vestres de Canarias: hongos, plantas y animales terrestres, gen), pp. 6–23. Brill and Backhuys, Leiden. 2009 (ed. by M. Arechavaleta, S. Rodrı´guez, N. Zurita and Solem, A. (1990) How many Hawaiian land snails are left? A. Garcıá), pp. 218–220. Gobierno de Canarias, Santa And what we can do for them. Bishop Museum Occasional Cruz. Papers, 30,27–40. Parent, C.E. & Crespi, B.J. (2006) Sequential colonization Triantis, K.A., Mylonas, M. & Whittaker, R.J. (2008a) Evolu- and diversification of Galapagos endemic land snail genus tionary species–area curves as revealed by single-island en- Bulimulus (Gastropoda, Stylommatophora). Evolution, 60, demics: insights for the inter-provincial species–area 2311–2328. relationship. Ecography, 31, 401–407. Parent, C.E. & Crespi, B.J. (2009) Ecological opportunity in Triantis, K.A., Nogue´s-Bravo, D., Hortal, J., Borges, P.A.V., adaptive radiation of Gala´pagos endemic snails. The Amer- Adsersen, H., Fernańdez-Palacios, J.M., Arau´ jo, M.B. & ican Naturalist, 174, 898–905. Whittaker, R.J. (2008b) Measurements of area and the Peake, J.F. (1981) The land snails of islands – a dispersalist’s (island) species–area relationship: new directions for an view. The evolving biosphere (ed. by P.L. Forey), pp. 247– old pattern. Oikos, 117, 1555–1559. 263. British Museum (Natural History), London & Cam- Triantis, K.A., Hortal, J., Borges, P.A.V. & Whittaker, R.J. bridge University Press, Cambridge. (2010) Patterns of endemicity in Macaronesia. Terrestrial Pokryszko, B.M. (1997) Lyropupa Pilsbry, 1900. Systematics, arthropods of Macaronesia: biodiversity, ecology and evolu- evolution and dispersal (Gastropoda: Pulmonata: Pupilloi- tion (ed. by A.R.M. Serrano, P.A.V. Borges, M. Boieiro dea). Genus, 8, 377–487. and P. Oromı´), pp. 45–64. Sociedade Portuguesa de En- Preece, R.C. & Gittenberger, E. (2003) Systematics, distribu- tomologia, Lisbon. tion and ecology of Balea (=Tristania) (Pulmonata: Clau- Triantis, K.A., Guilhaumon, F. & Whittaker, R.J. (2012a) siliidae) in the islands of the Tristan-Gough group. Journal The island species–area relationship: biology and statistics. of Molluscan Studies, 69, 329–348. Journal of Biogeography, 39, 215–231. R Development Core Team (2011) R: a language and envi- Triantis, K.A., Hortal, J., Amorim, I., Cardoso, P., Santos, A. ronment for statistical computing. R Foundation for Statisti- M.C., Gabriel, R. & Borges, P.A.V. (2012b) Resolving the cal Computing, Vienna, Austria. Azorean knot: a response to Carine & Schaefer (2010). Rabosky, D.L. & Glor, R.E. (2010) Equilibrium speciation Journal of Biogeography, 39, 1179–1184. dynamics in a model adaptive radiation of island lizards. Vega-Luz, R. & Vega-Luz, R. (2008) A new Hemicycla (Gas- Proceedings of the National Academy of Sciences USA, 107, tropoda: Helicidae) from Canary Islands. Malacologia Mos- 22178–22183. tra Mondiale, 61,24–26. Van Riel, P., Jordaens, K., Van Houtte, N., Martins, A.M.F., Whittaker, R.J. & Fernańdez-Palacios, J.M. (2007) Island bio- Verhagen, R. & Backeljau, T. (2005) Molecular systematics geography: ecology, evolution, and conservation, 2nd edn. of the endemic Leptaxini (Gastropoda: Pulmonata) on the Oxford University Press, Oxford.

Journal of Biogeography 40, 117–130 129 ª 2012 Blackwell Publishing Ltd R. A. D. Cameron et al.

Whittaker, R.J., Triantis, K.A. & Ladle, R.J. (2008) A general BIOSKETCH dynamic theory of oceanic island biogeography. Journal of – Biogeography, 35, 977 994. Robert A. D. Cameron has a long-term fascination with Whittaker, R.J., Triantis, K.A. & Ladle, R.J. (2010) A general the ecology, biogeography and evolutionary biology of land dynamic theory of oceanic island biogeography: extending snails. the MacArthur–Wilson theory to accommodate the rise and fall of volcanic islands. The theory of island biogeogra- Author contributions: R.A.D.C., R.J.W. and K.A.T. had the phy revisited (ed. by J.B. Losos and R.E. Ricklefs), pp. 88– original idea; F.G. and K.A.T. designed, performed analyses 115. Princeton University Press, Princeton, NJ. of data and wrote the methodological part; R.A.D.C., K.A.T. Workman, R.K., Hart, S.R., Jackson, M., Regelous, M., Far- and R.J.W. wrote the paper with the substantial contribution ley, K.A., Blusztajn, J., Kurz, M. & Staudigel, H. (2004) of the rest of the authors. Recycled metasomatized lithosphere as the origin of the enriched mantle II (EM2) end-member: evidence from the Editor: Melodie McGeoch Samoan volcanic chain. Geochemistry, Geophysics, Geosys- tems, 5, Q04008. doi:10.1029/2003GC000623. Zuur, A., Ieno, E.N., Walker, N., Saveiliev, A.A. & Smith, G. M. (2009) Mixed effects models and extensions in ecology with R. Springer, New York.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article: Appendix S1 Datasets used for the land snails of eight oceanic archipelagos (Tables S1 & S2). Appendix S2 Island groupings considered for the land snails of eight oceanic archipelagos. Appendix S3 Supplementary analyses and results (Tables S3 & S4).

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing ﬁles) should be addressed to the authors.

130 Journal of Biogeography 40, 117–130 ª 2012 Blackwell Publishing Ltd